FASTA giving bad results?

Howdy,

I still a new user, and I am simulating a large log-periodic slot array using magnetic current simulation in IE3D. I am wanting to expand the current design to cover a larger frequency range, but doing so creates too many variables to be ran in a reasonable amount of time.

I found that the FASTA feature in IE3D could significantly reduce simulation time, but it appears that the results are completely inaccurate! I varied the "minimum unknowns" in the FASTA settings and attached a graph comparing results.

Can anyone tell me why the results are so inaccurate? Does FASTA have any limitations other than listed in the user manual?

Settings used:
Fmax=20
Ncells=15 (should this be bigger/smaller for good results?)
AEC=0

Thanks in advance for any help you can give me.

-Andy

Hi anybody
I have same problem with FASTA in simulation of an Aperture coupled Patch Array Antenna with 64 elements.

Regards
KMPA

Hi:

Please send the problem files to support@zeland.com. We will try to study them and try to improve FASTA. It is still new. Also, all fast algorithms will have limitations. From what I see, FASTA is the one has the least limitation. We have tested it for different kinds of structures. Overall, it is quite stable for many structures. However, there are exceptions. There should still bel much room to improve it. We welcome feed back from users. Thanks!

Can you suggest an alternate matrix solver to shorten my simulation times, or at least enable them to be ran on a computer with only 2G of ram?

FYI, I plan to expand the above design which will increase the cells 6X (without AEC) so time-savings is very important.

-Andy

Hi, bazookazuz:

You can try to use the GEMS-F iterative solver. Your structure has 8076 unknowns. It may require about 540 MB RAM for the SMSi solver. If you have 2 GB, you should be able to solve it using SMSi solver quite easily. It may take 10-20 minutes per frequency point. I noticed that you define coarse frequency points. For most structures, you may consider putting more points (100-500). Putting more points will not slow down the simulation because IE3D nromally can finish it using about 6-20 points. For this particular structure, I am afraid it may take more than 20 points because there are so many resonances from 1 to 20 GHz.

If you want to increase the size of the strucutre and SMSi matrix does not have enough RAM for it. You can use GEMS-F iterative solver. It can reduce the RAM or may speed up the simulation. I am running your structure with 8076 unknowns. It does not seem to me the GEMS-F can be faster than the SMSi significantly. However, I can see it reduces the RAM requirement from 540 MB to below 100 MB. GEMS-F solver should yield accurate results as long as the iteration can converge.

Best regards.

Added after 1 minutes:

By the way, your log array is a slotted log array because z = 0 is the ground plane. Your polygons are on z = 0 and IE3D is using m-current modeling for the slot structure. Just want to make sure.

Added after 5 minutes:

Sorry. It was my mistake. The polygons are on z = 1.6 and z = 1.6 is a ground plane. The array is a slotted log array. Regards.